Avalable onlne at www.scencedrect.com Energy Proceda 37 (2013 ) 952 960 GHGT-11 CO 2 absorpton wth membrane contactors vs. packed absorbers- Challenges and opportuntes n post combuston capture and natural gas sweetenng Karl Anders Hoff a * and Hallvard F. Svendsen b a SINTEF Materals and Chemstry, Trondhem NO-7465, Norway b Norwegan Unversty of Scence and Technology, Trondhem, Norway Abstract The present work presents a case study where membrane contactors are compared wth absorpton towers for post combuston CO 2 capture and for natural gas sweetenng. Smulatons are performed usng a absorber model whch has been valdated wth experments n a membrane contactor setup. The desgn of the membrane contactors s made wth emphass on the constrants n gas and lqud sde pressure drop and sze lmtatons of membrane modules, n addton to the common desgn crtera for ndustral absorbers. Results show that absorber sze may potentally be reduced by 75%, gven that lqud s flowng on the shell sde of the membrane unts. Natural gas sweetenng s a more vable opton than post combuston capture due to the potentally large gas sde pressure drop and the need for many membrane unts n parallel n the latter case. 2013 The Authors. Publshed by by Elsever Ltd. Ltd. Selecton and/or peer-revew under responsblty of GHGT of GHGT Keywords: Type your keywords here, separated by semcolons ; 1. Introducton Membrane contactors represent a novel absorber desgn that can offer several mprovements over conventonal packed columns. These are manly related to the operablty and to the sze and volume of the equpment. The overall process topology s the same as n a conventonal absorpton desorpton cycle and the energy requrement for CO 2 capture s stll a result of solvent performance and process optmzaton (Fgure 1). * Correspondng author. Tel.: +47 982 43482 E-mal address: karl.a.hoff@sntef.no. 1876-6102 2013 The Authors. Publshed by Elsever Ltd. Selecton and/or peer-revew under responsblty of GHGT do:10.1016/j.egypro.2013.05.190
Karl Anders Hoff and Hallvard F. Svendsen / Energy Proceda 37 ( 2013 ) 952 960 953 Even f membrane contactors have receved sgnfcant nterest n the lterature, see revew by Favre and Svendsen (2012), thorough evaluatons of the technology n a realstc process envronment are stll lackng. Ths requres a fundamental understandng of membrane contactor performance combned wth knowledge of optmal process desgn for absorpton/desorpton cycles. The latter has receved sgnfcant attenton durng the prevous 5-10 years, especally wthn the feld of post combuston CO 2 capture. The current work presents results from a case study of membrane absorbers compared wth conventonal absorbers. The current work s based upon three dfferent gas treatment optons, all of sgnfcant ndustral mportance: 1. Post combuston CO 2 capture from offshore gas turbnes 2. Sweetenng of natural gas wth 10% CO 2 to ppelne specfcaton of 2.5% 3. Same as 2 wth LNG specfcaton of 50 ppm Overhead condenser Treated gas Conc. acd gas Absorber Lean solvent cooler Desorber Reflux Membrane Absorber Membrane Desorber Lean/Rch HX Feed gas Rch solvent Lean solvent Reboler Fgure 1 Smplfed absorpton desorpton cycle showng how membrane contactors may replace the packed columns 2. Membrane contactors and packed absorbers 2.1. Prncple The prncple of membrane gas absorpton s llustrated n Fgure 2. A mcroporous membrane works as a non-selectve barrer between the gas and the solvent. As long as the mass transfer resstance s located on the lqud sde and the membrane materal s not wetted by the solvent, the mass transfer resstance of the membrane wll be lmted. Ths requres the pore sze to be larger than the mean free path of dffusng gas. Wth a typcal aqueous chemcal solvent t s mportant that the membrane materal s hydrophobc. Materals used n ths applcaton are polymers lke PTFE or polypropylene and the membranes are typcally made as hollow fber bundles or sheets wth fbers of 0.5-3mm ID. The man postve features of membrane contactors are:
954 Karl Anders Hoff and Hallvard F. Svendsen / Energy Proceda 37 ( 2013 ) 952 960 No floodng, entranment, channelng and foamng: The presence of a membrane between gas and lqud prevents these phenomena from occurrng, as opposed to packed columns where they may drastcally reduce separaton performance. No senstvty to moton: Even mnor tltng of an absorpton tower wll reduce performance. Ths s an mportant ssue when operatng on an offshore unt. In post combuston operaton on e.g. gas turbne exhaust, the low specfc lqud load may not utlze the packng surface area completely. Membrane contactors can therefore represent a sgnfcant mprovement n ths case. Hgher specfc area/reduced sze of equpment: Packed columns typcally offer an actve nterfacal area of 100-250m 2 /m 3. Hollow fber membranes unts can be stacked n prncple wth very hgh specfc surface area. It s often clamed >1000 m 2 /m 3. However, the packng densty of fbers n a membrane contactor s lmted by pressure drop constrants both for the gas and the lqud flow. Secondly, n a system wth mass transfer followed by chemcal reacton wthn a thn concentraton boundary layer near the membrane surface, the membrane porosty reduces the true mass transfer area to some extent. Lqud sde mass transfer: When lqud s flowng nsde the hollow fbers n undsturbed, lamnar flow, the mass transfer coeffcent s low. Mass transfer can be sgnfcantly enhanced by ntroducng lqud mxng ponts, thus reducng the resdence tme of lqud elements on the nterface. (Hoff et al. 2006). The smplest way of ntroducng lqud mxng s to pass the lqud on the shell sde and gas on the tube sde. Membrane tube layer Spacer Gas flow Counter /cross current Fber pottng Lqud flow Fgure 2 Prncple of membrane contactors 2.2. Desgn consderatons Desgn of absorpton columns and membrane contactors are subject to some of the same constrants as well as some dfferences. The fundamental relatonshp between capture rate, gas flow and lqud crculaton s vald for both: Q L n yco Q P G RTG 2 ( ) m rch lean (1)
Karl Anders Hoff and Hallvard F. Svendsen / Energy Proceda 37 ( 2013 ) 952 960 955 where m s the solvent concentraton, s the CO 2 loadng, y CO2 s the mole fracton of CO 2 n the feed gas, s the CO 2 recovery. Q G s the volumetrc flow of gas and Q L s the volumetrc lqud crculaton rate. The rch and lean loadngs are crtcal parameters for optmal operaton of an absorpton process. rch s gven by the need to acheve a hgh approach to equlbrum n the absorber bottom. For MEA absorbers n post combuston operaton >90% approach to equlbrum s common, correspondng to values of >0.45 mol CO 2 /mol MEA. Lean loadng s typcally 0.25 n a MEA desgn optmzed for low reboler duty. For natural gas sweetenng wth MDEA based solvents, the approach to equlbrum s usually lower and also less crtcal to the energy requrement. 60-80% of equlbrum s more common. MDEA based solvents can normally be regenerated to almost zero loadng by a combnaton of flash stages and strppng. The man pont s that rch and lean loadngs are more or less fxed for a gven separaton problem and that the lqud flow, Q L, as well as the lqud to gas rato Q L /Q G wll be the same n a membrane contactor system as n a packed absorber. When desgnng a conventonal absorpton tower the followng addtonal constrants are mportant: For post combuston capture, the total pressure drop on the gas sde should not be hgher than 0.1 bar, n order to utlze a conventonal blower for the flue gas. Hgher pressure drop wll also lead to hgher dutes on the blower, whch s the largest consumer of electrc power n the plant. The superfcal gas velocty should be as hgh as possble to mnmze tower dameters, but s lmted by floodng n the column and the desre for low pressure drop. Gas veloctes off 2-3 m/s are typcal n post combuston capture. The packng should be wetted by lqud dstrbuted over the whole tower cross secton. For a typcal natural gas exhaust case, the lqud load wll be around 10 m 3 /m 2 h. Ths s n the low range of lqud loads where ncomplete wettng sometmes may arse. The packng bed must be splt n several sectons for lqud redstrbuton. For membrane contactors, the constrants are the followng: The constrant n terms of gas sde pressure drop s the same as for towers. Gas sde pressure drop wll ncrease as the packng densty/the specfc surface area ncreases. Also, f gas s flowng on the tube sde, long membrane passes wll ncrease the total pressure drop. The lqud entry pressure represents an upper lmt at whch penetraton nto the membrane pores should not occur. The lqud sde pressure drop wll therefore be constraned by the maxmum fber length allowed. Ths can partally be compensated by utlzaton of head pressure n vertcally orented unts. There are lmtatons related to the physcal dmensons of the membrane modules. Dependng on the membrane manufacture/spnnng process, the length of the hollow fber cannot be more than a few meters. Commercal hollow fber membrane modules manufactured today typcally stll have dameters less than 1m (Marquez and Brantana, 2006). n 3. Model development and valdaton Earler, a rgorous modelng and desgn tool for membrane absorbers was developed (Hoff et al. 2004). The model was developed over several years and valdated vs. more than 500 experments n membrane absorbers both n bench and plot scale.
956 Karl Anders Hoff and Hallvard F. Svendsen / Energy Proceda 37 ( 2013 ) 952 960 The model s based upon the soluton of the 2-dmensonal equatons for conservaton of mass and energy n a membrane hollow fber. Detals are gven n Hoff et al. (2004) but the model has gone through some sgnfcant further development recently. These mprovements are mportant n terms of desgnng large scale unts. Improved thermodynamc models for the MEA and MDEA/Pperazne based systems Development of a desgn model based upon mass transfer coeffcents for the lqud sde Implementaton of lqud mxng geometry Smulaton of lqud flow on the shell sde Improved models for gas and lqud sde pressure drop Better mass transfer models, ncorporatng effects of partal lqud penetraton, as well as Knudsen dffuson n the membrane pores. The followng descrpton hghlghts some mportant aspects of the model relevant for the current work: 3.1. Mass transfer flux The transfer flux across the membrane s modelled usng the smple resstance n seres approach, ncludng the gas flm and the membrane mass transfer coeffcents. 1 eq N = (f - f ) x N 1 1 ZRT g k g, km (2) eq where f and f are the fugactes of the transferable components, actually n the gas phase and at equlbrum wth the lqud. The last term n Eq. 2 s the convectve flux. The gas flm coeffcent, k g, s gven by a correlaton based upon lab-scale measurements n a gas controlled regme, usng SO 2 absorpton n NaOH as model system. The membrane mass transfer coeffcent, k m, s gven by k m D p R ln R R o (3) where s the membrane porosty, s the membrane tortuosty whle R and R o are the nner and outer rad, respectvely. D p s a combned term accountng for molecular dffuson and possble contrbutons from (a) Knudsen dffuson (facltated by small pore dameters/low pressures) and (b) partal lqud penetraton n the membrane pores: D p 1 x l 1 1 1 1 xl D D D m Kn m, lq (4)
Karl Anders Hoff and Hallvard F. Svendsen / Energy Proceda 37 ( 2013 ) 952 960 957 To reduce the computatonal tmes, a smplfed model has been formulated. Ths s also requred n order to smulate lqud on shell sde. In ths model, equaton (2) s reformulated, ncludng a mass transfer coeffcent for the tube sde, n the followng case wth lqud nsde the fbers: N = ZRT g 1 1 1 H k g k Ek, k,, m 0 Ek, l * (f - f ) x N (5) The tube sde coeffcent can be gven by the Graetz-Nusselt soluton for the local mass transfer coeffcent (Cussler, 2009). When the hydrodynamcs of the lqud phase are descrbed by a physcal mass transfer coeffcent, an enhancement factor s requred to account for the effects of chemcal reacton. Here s used the approach by De Coursey et al. (1982), combned wth the enhancement factor for an nstantaneous reversble reacton, E, based upon Olander (1960). 3.2. VLE model An equlbrum model for the CO 2 /amnes/water system s ncluded, descrbng the complex chemstry of the CO 2 /amne/water system. Ths s based upon the Desmukh-Mather approach, whle non-dealtes for the gas phase are calculated by the Peng-Robnson Equaton of state. Ths s partcularly mportant for the current work snce both the compressblty factor and the fugacty coeffcents devate sgnfcantly from 1. Compared to the orgnal Desmukh Mather model (Weland et al. 1993), a modfed Debye- Hückel model was adopted, smlar to the paper by Kuranov et al. (1997). Ths gave a slghtly better ft to the data. The model s developed for the solvents MEA/water and MDEA/Pperazne/water. 3.3. Model valdaton Fgure 3 shows model predctons of the absorpton flux n a lab-scale membrane absorber unt operatng on 30wt% MEA. The expermental condtons span a wde range n terms of lqud velocty, temperature, gas phase CO 2 concentraton and lqud loadng. Fgure 3 Model predcton vs. expermental absorpton rates, 30wt% MEA.
958 Karl Anders Hoff and Hallvard F. Svendsen / Energy Proceda 37 ( 2013 ) 952 960 4. Present study 4.1. Reference smulaton wth packed tower A complete process smulaton was bult for the post combuston and natural gas sweetenng cases. For the post combuston case, the SINTEF/NTNU software CO2SIM (Tobesen and Schumann-Olsen, 2011) was used, whle the natural gas cases were smulated n Protreat (Optmzed Gas Treatng Inc.). Results and absorber dmensons are gven n Table 1. The gas and lqud flows were used as nput to the membrane contactor model to desgn an absorber system for the same purfcaton targets. Membranes smulated n ths work are mcroporous PTFE hollow fbers of smlar type as n Hoff et al. (2004). For post combuston the membrane modules are arranged n a cross flow pattern. The rectangular modules have a fber length of 3m, 3m wdth and 1m depth. The hgh pressure modules used for natural gas treatment are desgned as cylndrcal cansters lke a shell and tube heat exchanger, wth maxmum dameter 2-2.5m and maxmum length 5m. These dmensons are not related to any specfc membrane manufacture process and are only used as a bass. Further work s requred n order to nvestgate the lmtatons n module sze. Results from the desgn of the membrane absorber unts are shown n Table 2. It s seen that the total absorber volume s reduced by typcally 75% compared to towers. Ths requres lqud flow on the shell sde, as tube sde flow provdes too low lqud sde mass transfer coeffcents. However, for the post combuston case, ths comes at a prce of too hgh pressure drop on the gas sde. An addtonal challenge wth post combuston s the large volumetrc gas flows, requrng a hgh number of membrane modules n parallel. For natural gas treatment, gas flow s lower and lqud flow s hgher, and the separaton can be acheved wth only 3 membrane modules n parallel. Table 1 References cases wth packed absorbers, based upon smulatons n CO2SIM and Protreat Treated gas Post combuston (PC) Natural gas (NG) Natural gas (NG) Specfcaton 90% capture 2.5% CO 2 50 ppm Solvent 30% MEA 49% MDEA 45% MDEA + 5% Pz Tot pressure, feed gas (kpa) 100 7 000 7 000 CO 2 content, feed gas (vol%) 3 10 10 CO 2 recovery (%) 90.00 77.25 99.96 Lean loadng 0.21 0.0029 0.011 Rch loadng 0.45 0.361 0.314 Lqud flow (m 3 /h) 1 183 1 429 1 944 Gas flow (actual, m 3 /h) 1 311 619 8 381 8 381 Column Dameter (m) 13.6 4.4 4.6 Packng heght (m) 18.6 14.0 15.0 Packng volume (m 3 ) 2 702 208 246 Packng materal Mellapak 250.Y 2.5" Nutter Rng 2.5" Nutter Rng Superfcal gas velocty (m/s) 2.5 0.16 0.14 Gas sde pressure drop (kpa) 3.7 2.4 2.2
Karl Anders Hoff and Hallvard F. Svendsen / Energy Proceda 37 ( 2013 ) 952 960 959 Table 2 Results from smulatons of membrane contactor system Case Post combuston Post Combuston NG, ppelne NG, ppelne. NG, LNG Solvent 30% MEA 30% MEA 30% MDEA 30% MDEA 30% MDEA 5% Pz Lqud flow mode Shell sde Tube sde Shell sde Tube sde Shell sde Module wdth (m) 3.0x1.0* 3.0x1.0* 2.10 2.40 2.40 Fber length (m) 3.0 3.0 5.0 4.5 5.0 # modules n parallel 25 40 1 2 1 # modules n seres 3 5 3 7 5 Pressure drop, gas (kpa) 49.20 5.7 2.32 10.54 1.03 Pressure drop, lq. (kpa) N/A 37.1 N/A 111.6 N/A Superfcal gas vel. (m/s) 1.43 0.93 0.22 0.07 0.10 Total packed volume (m 3 ) 675 1800 52 285 113 Volume rato vs. tower 0.25 0.67 0.25 1.37 0.46 (*Cross flow modules, W=3m, H=3m, D=1m) 5. Conclusons The results from the present study show that sgnfcant savngs are possble n terms of absorber sze when replacng a conventonal column wth a membrane contactor. Ths requres lqud to flow on the shell sde to enable splttng and mxng of lqud elements, smlar to the lqud mxng provded by structured or random tower packngs. Membrane contactors may provde sgnfcant mprovements n offshore CO 2 capture, both from gas turbne flue gas and n sweetenng of natural gas. Pressure drops, both on the lqud and gas sde, represent constrants that must be ncluded n the desgn of a membrane contactor system. Modularty and sze lmtatons of membrane systems also represent a challenge. Ths s especally the case for post combuston, where there wll be a need for a large number of unts n parallel and n seres, as opposed to conventonal absorpton towers. It s thus uncertan whether membrane contactors can offer an advantage n ths case. Acknowledgements Ths publcaton has been produced wth support from the BIGCCS Centre, performed under the Norwegan research program Centres for Envronment-frendly Energy Research (FME). The authors acknowledge the followng partners for ther contrbutons: Aker Solutons, ConocoPhllps, Det Norske Vertas, Gassco, Hydro, Shell, Statol, TOTAL, GDF SUEZ and the Research Councl of Norway (193816/S60). Ths publcaton s based on the results f Petromaks program. The author(s) acknowledge the partners: Statol, Total, Gassco, Petrobras and the Research Councl of Norway (200455/S60) for ther support.
960 Karl Anders Hoff and Hallvard F. Svendsen / Energy Proceda 37 ( 2013 ) 952 960 6. References Cussler, E.L."Dffuson", Cambrdge Unversty Press, 2009 De Coursey WJ. "Enhancement Factors for Gas Absorpton wth Reversble Reacton", Chem. Engng. Sc. 1982 (37), pp.1483-1489 Favre E. and H.F. Svendsen, "Membrane contactors for ntensfed post-combuston carbon doxde capture by gas-lqud absorpton processes", J. Memb. Sc. 2012; 407-408, 1-7 Hoff KA., Bjerve F., Mejdell T., Dndore V., Julussen O. and H.F. Svendsen, wth enhanced mxng geometry - characterzaton, modelng and process mp Control Technologes, GHGT-8, Trondhem, 2006 Hoff, KA., Julussen, O., Falk-Pedersen, O. and H.F. Svendsen, "Modelng and Expermental Study of Carbon Doxde Absorpton n Aqueous Alkanolamne Solutons Usng a Membrane Contactor", Ind. Eng. Chem. Res. 2004 (43), pp. 4908-4921 Kuranov, G, Rumpf B, Maurer B. and N. Smrnova, "VLE modellng for aqueous systems contanng methyldethanolamne, carbon doxde and hydrogen sulfde", Flud Ph. Eq., 1997, Volume 136, Issues 1 2, pp. 147-162 Marquez, J.and M. Brantana, "Hgh-capacty gas membrane reduce weght, cost", Ol & Gas J. July 24. 2006 Olander, DR. Smultaneous mass transfer and equlbrum chemcal reacton. A.I.Ch.E. Journal. 1960;6:233-239 Tobesen, A. and Henrk Schumann-Olsen, Obtanng optmum operaton of CO 2 absorpton plants, Energy Proceda, (4) 2011 pp 1745-1752 Weland, R.H., Chakravarty, T., Mather, A.E., Solublty of Carbon Doxde and Hydrogen Sulfde n Aqueous Alkanolamnes, Ind. Eng. Chem. Res. 32, 1993, pp 1419-1430